The AI coding apocalypse
www.businessinsider.com
In 2023, not long after ChatGPT made generative AI mainstream, a poll on the anonymous workplace forum Blind asked, bluntly, whether young software engineers were "fucked." Some 42% of the more than 13,000 respondents picked the response "Yes? U guys are pretty much fucked."This past October, Sundar Pichai proudly announced on an earnings call that AI was writing more than 25% of new code at Google. Mark Zuckerberg has said that Meta will build an AI engineer to write code. Salesforce CEO Marc Benioff announced a hiring freeze for engineers in 2025, saying AI had increased productivity by 30% and then news broke that Salesforce planned to lay off 1,000 workers. (It's still hiring salespeople for AI-powered products.) Stripe intends to cut some engineers while also growing its overall head count this year.All of this raises the question of what junior engineers will take on if some basic tasks become automated. Some product managers have speculated that AI will increasingly take on some technical coding tasks and circumvent their need for engineers. Overall, job postings for software engineers on Indeed are at a five-year low.Are engineers really coding themselves into obsolescence?AI is knocking down the career ladder by doing more of the coding work of entry-level engineers, but, at least for now, the increased coding output from AI is also increasing the demand for and value of experienced, creative developers to interpret and put the AI's work to good use.While many obituaries have been written to mourn the death of coding, engineering is more than writing code: It requires creative thinking to solve problems and expertise to read code. As it is now, AI isn't an original thinker."AI can't support what it doesn't know," says James Stanger, the chief technology evangelist at CompTIA, a nonprofit trade association for the US IT industry. "I still don't think that it is something that can fully replace a good developer." He adds, though, that "if a developer is not creative, then you can replace them very easily."oftware engineering has been around since the 1960s, but hiring boomed in the '90s with the dot-com era. Coding boot camps became common in the 2010s as the demand for engineers outpaced the supply. According to the Bureau of Labor Statistics, 1.9 million people worked as software developers, quality-assurance analysts, and testers in 2023. The bureau projected that the industry would grow by 17% from 2023 to 2033, outpacing the national average of 4% for all jobs.An analysis from CompTIA found that the rate of job postings for software engineers fell by 50% from January to December 2023, recovering slightly by the end of 2024. While posts for jobs across tech, finance and accounting, and marketing, communications, and creative roles all fell as well, the dip for software engineers was the sharpest. But CompTIA says the reason for the fall wasn't clear. Tech companies including some that acknowledged overhiring during the pandemic laid off thousands of workers in 2022 and 2023, with many citing economic uncertainty.But there's a widening divide within software engineering regarding experience level. CompTIA found that the proportion of open software engineering roles seeking entry-level workers had dropped since January 2023, to just over 20% from nearly 30%, while job postings for those with seven years of experience or more increased to make up nearly 40% of the open roles, up from just over 30%.The experienced engineers I talked to seemed confident that AI wouldn't come for the jobs anytime soon.Jeremy Chua, a software engineer for the AI Lab at the venture firm Georgian, turns to chatbots when he hits issues with coding. He may prompt ChatGPT or Claude to cull answers from the depths of Google and Stack Overflow, a Q&A site for programmers, or to help him write in coding languages he's less fluent in. Chua, who has more than a decade of experience, says he was skeptical about whether gen AI could help him at work. He says that now he can sometimes complete projects that would have taken a week in a day or two, and he thinks of the chatbots he uses as coding partners. "It's not like it will replace me it augments the way that I work," Chua tells me.Caleb Tonkinson, an engineer at a clinical AI company called SmarterDx, tells me that AI is changing programming through two paths: "I can deliver the same thing faster, or I can deliver something better in the same period of time." He views AI as similar to other tech tools that became available to engineers except more exciting as it advances rapidly. "There have been tons of tools for 20 years now" to debug software, generate code, or evaluate code, he says. "Your best companies and best software engineers are almost always leveraging those tools."Cody Stewart, a principal software engineer at the software company CallRail, says he doesn't use gen AI for everything at work but might use it to get answers to "stupid questions" that he could spend a long time looking for on Google or Stack Overflow. He began using chatbots at work in 2022. "I read something that was like, you either learn to adopt new tools and figure out how they can enhance your day-to-day life and you stay with the times, or people are going to outlevel you," he said. "I saw that and thought I should probably give this a shot."While more-experienced engineers are optimistic about AI, young engineers have more reason to worry.The startup Cognition AI last year widely released an AI-powered software engineer called Devin, designed to work on bugs and small feature requests. In a December video, it described it as "a junior engineer" who "works best with a great manager." Cognition AI and its CEO, Scott Wu, did not respond to questions about whether it's meant to replace engineers or reduce the number the companies need.Jayesh Govindarajan, a Salesforce executive vice president focused on AI, told my colleague Ana Altchek that the company was building "a system that can pretty much solve anything for you" but "just doesn't know what to solve," making knowing how to code less important. "I may be in the minority here," Govindarajan said, "but I think something that's far more essential than learning how to code is having agency."Alexander Petros, a freelance open-source software engineer, is an AI holdout; he tells me he doesn't use generative AI to code. "I do worry that because AI is in many ways doing things that you used to hire junior developers to do, it does remove the ladder upon which junior developers would try to do those things, make those mistakes, and then learn," he says. Petros says he tried ChatGPT but found that the code could be clunky. If something in that code breaks, humans may not know how to fix it. "The process of producing code with LLMs, for the foreseeable future, is almost entirely distinct from the process of producing good software systems that last for a very long time," he says. Plus, using AI to solve problems means he may not learn how to get through those roadblocks on his own.Chatbots lack creativity that's where engineers, especially those who have been doing the job for a while, have an in-demand advantage. Stanger says he hopes companies use AI not as justification for cutting back on engineers but as a way to help them "get deeper into this code and get more creative."Stanger says that treating engineers as a faucet that can be turned on and off as a business needs, or even replaced with AI, is likely to backfire in the long term. "If you've got toxic companies that are interested in that binge-and-purge, on-and-off hiring of developers, I'm not sure they're going to create very good products," he says.People have long panicked that technology will take their livelihood. But even as automation eliminates some jobs, tech often creates a demand for new roles; most people today are working jobs that didn't exist before 1940.The wholesale elimination of software engineers likely won't come to fruition in the near future, but the picture for more-experienced engineers is brighter. In the best-case scenario, AI will mean they get more time to flex their muscles and solve deep problems.Amanda Hoover is a senior correspondent at Business Insider covering the tech industry. She writes about the biggest tech companies and trends.
0 Yorumlar
·0 hisse senetleri
·73 Views